MPU Meeting Tuesday 12th April 2011

Software Build Farm

Graham carried out some stress tests of the build farm using matlab. This is a 2GB source package which generates packages that are nearly 1GB in size. This revealed that the AFS caches were far too small on the various servers. To improve matters the build servers have been reinstalled with much larger AFS caches. The master server, ardbeg, still needs to be reinstalled with a larger AFS cache.

Stephen has also made a few tweaks to the code to reduce the amount of unnecessary copying of source files which should help with big packages. Those changes have not yet been shipped, hopefully they will go out soon.


The PXE installer now uses a native SL6 install image for both architectures.

We now have two SL6 build hosts - leeds for i386 and kilmartin for x86_64. They will need to be reinstalled as proper dice machines at some point but it's a good starting point for other units needing access to SL6 for development work.

Most of the work to transfer MPU SL6 resources from the inf to dice layers has been done.

There are a few medium priority components still to be ported. After that the next important stage is to add the server support. The lack of nagios support is going to be a bit of a nuisance. We should offer to help Inf Unit with the porting of the nagios component to the new LCFG build tools.

We should start testing SL6 on our various desktop models. Stephen will talk to Carol and see if she has time to do some testing.

Alastair will publicise the SL6 upgrade to everyone in the School.

AFS automation project

Chris has got a test cell installed - - this has a DB server and two file servers. He has been playing with creating, destroying and moving volumes to learn how it all works. For time monitoring purposes we will count this stage of the project as personal development.

Craig has added another desirable facility, we would like to have support for dynamic quotas. This will go on the list, it needs to be prioritised.

Miscellaneous Development

sleep component
The idea of checking for running cron jobs was discussed. Chris has been looking at the Proc::ProcessTable module. We decided that the best approach was to read the PID from the cron daemon pid file and then scan the process table for any process with that PPID. That should catch all normal running cron jobs.

Memory for brendel
Alastair has ordered more RAM for brendel so we can do more testing of storing all the LCFG server data files entirely in shared-memory.


Server Reboots
We have now done all the MPU server reboots.

Still need to look at the (possible) problems with the local RAID controller/disks. We might also want to look at applying any BIOS upgrades.

Packages volumes
The Services Unit need to move the packages volumes to a new file server with more space. We will need to coordinate with them so that we turn off the refreshpkgs daemon whilst the move is happening.

This needs to be switched to using dhcp before we can move it to the AT machine room.

kernel cpt
The new support for listing kernel modules associated with a source package is now available. This should reduce the number of reboots required after an install or kernel upgrade.

AFS cache size
The default AFS cache size has now been changed to 8GB.

DR server
Stephen has setup a mirror of the LCFG install ISO images on sauce. He has also added a mirror of the SL6 repositories and tidied the profile so that all the DR config is in LCFG headers. We need to document how to restore the various services in the event of disaster. We also need to switch sauce to the stable release.

updaterpms patches
Alastair will take a look at Stephen's patches for updaterpms.

This Week

  • Alastair
    • SL6 - server functionality
    • SL6 evangelism
    • Development meeting activity page for SL6
    • Add memory to brendel and test LCFG compilation (both with/without shm)
    • Convert satablade to using DHCP
    • Look at Stephen's patches to updaterpms component
    • Consider personal development topics

  • Chris
    • AFS automation project
    • Development meeting activity page for AFS automation
    • Consider personal development topics

  • Gordon

  • Stephen
    • SL6
    • Ask Carol to do desktop testing
    • Development meeting activity page for PkgForge
    • stabilise sauce
    • Consider personal development topics

-- AlastairScobie - 12 Apr 2011

Topic revision: r4 - 19 Apr 2011 - 13:21:45 - StephenQuinney
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies